Page 3 of 3 FirstFirst 123
Results 41 to 44 of 44

Thread: Let's talk Neural Network.

  1. #41
    Non of your ve tables are used. The NN uses the virtual ve aka nn to adjust fuel depending on where the cams are. You need to use the nn trainer to tune it properly. If that was a concern anyhow.

  2. #42
    Advanced Tuner
    Join Date
    Aug 2009
    Posts
    224
    Quote Originally Posted by luciddeath View Post
    Well on all of my supercharged pentastar setups i tune the NN. Feel it isn't really tuned if you turn it off on a 6. Lol
    So you are paying for the tokens and using the NN trainer, or just leaving it on and tweaking the weights and such? I can see the extra 100 bucks being worth the trainer on a supercharged setup.

  3. #43
    I do pay the trainer and use it only as a last resort if im having issues with the injectors. Fine tuning injectors if getting the pw's down perfect isn't enough.

  4. #44
    Tuner in Training
    Join Date
    Jun 2020
    Posts
    34
    NN will work fine if you rescale your tables for the additional atmosphere when going FI and you have the injector flow data correctly translated for HPtuners. The biggest issue is the injector resolution in the HPtuner table is really small compared to the whole injector range, so if you have it off by a bit the end result if your calculation/line could be off by 20%_30% either way

    The problem with turning it off on VVT/VVL cars is there is not multiple VE tables to account for all the different cam configurations you could possibly have, the modifiers can get you close bit at part throttle you will most likely have transient lag in fueling.

    The biggest pain in the whole process are the airflow model and coefficients that cap it based on throttle body air inlet calculated. This forces you to falsely adjust that airflow and then build around PRatio. Weights in the NN should reflect accordingly.